jina-ai/embedding-inversion-demo

Embedding Inversion via Conditional Masked Diffusion: recover original text from embedding vectors using parallel denoising. Live demo + training pipeline + technical report.

36
/ 100
Emerging

This project helps security professionals and data privacy officers understand the reversibility of text embeddings. It takes an embedding vector, which is a numerical representation of text, and reconstructs the original words from it. This tool can be used by those concerned with data leakage or the potential for sensitive information to be recovered from supposedly anonymized text embeddings.

Use this if you need to assess the vulnerability of text embeddings to reconstruction attacks or demonstrate that seemingly irreversible text data can be recovered.

Not ideal if you are looking for a tool to generate text from scratch or to improve the quality of your text embeddings.

data-security privacy-assessment text-anonymization vulnerability-testing
No Package No Dependents
Maintenance 10 / 25
Adoption 7 / 25
Maturity 11 / 25
Community 8 / 25

How are scores calculated?

Stars

40

Forks

3

Language

Python

License

Apache-2.0

Last pushed

Mar 07, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/jina-ai/embedding-inversion-demo"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.